Liquid cooling can be a cost-effective thermal management solution, sometimes the only feasible solution, in some practical computing applications such as high-density servers/racks and high-performance compute (HPC) servers/racks. HPC, for instance, uses high performance processors such as CPU and GPU processors that generate a large amount of heat during operation. Thermal management for these high power density processors and servers thus becomes a challenge.
There are few attempts to solve this thermal management challenge. For liquid cooling solutions using a cold plate, most existing solutions are single cold plates with supply and return hoses assembled together. Assembling a single cooling element onto a processor is easy, but once there are multiple cooling elements assembled in a cooling loop, installing the cooling loop on the processors is very difficult and risky. If the cooling loop has multiple cooling elements, it can require multiple operators to work together to complete assembly, which increases cost and the probability of errors.
There is no mature or effective solution for solving this problem. In terms of leakage prevention, existing solutions focus on non-conductive working fluid, negative pressure system, and leak detection. But all these approaches inefficient.